Webvan → Instacart, DoorDash, Amazon Fresh
Kozmo.com → Postmates, Uber Eats, Gopuff
Boo.com (fashion) → Farfetch, Net-a-Porter, ASOS
Broadcast.com → YouTube, Netflix, Twitch
The dot-com bubble didn’t prove the internet was a fad — it proved the internet was inevitable, but the valuations assumed adoption would happen in 2 years instead of 15–20. To me it feels like the AI inevitability will be much quicker.
Based on what? We're only seeing linear improvements for increasing spending. There's no new algorithm ideas on the horizon, just more and more hardware, in the hopes that if we throw enough RAM and CPU at the problem, it will suddenly become "AGI."
No one has their eye on power budgets or sustainability or durability of the system. The human brain has such a high degree of energy efficiency that I don't think people understand the realities of competing with it digitally.
The main problem "AI" seems to solve is that humans get bored with certain tasks. The language models obviously don't, but they do hallucinate, and checking for hallucinations is an exceedingly boring task. It's a coffin corner of bad ideas.
Internet did not have enough devices to reach people. At the height of 2002 only a fraction of people worldwide had an already expensive computer and an internet to go with it.
I ran a e-commerce startup from 2005-2010. Having access to demand is a thing.
Today everyone has access in their pockets. Go to small city in Africa, India and China, and observe how they use AI. See how perplexity has put AI answers in hundreds of millions of people's hands before Google in a matter of months.
Forgive me for saying — but "Based on what" for comparing accelerated adoption between 2005 and 2025 — is discarding many huge elephants in the room, starting with that small thing you're reading this in your hand with, and the invisible thing that's sending you this comment.
The LLM algorithms seem pretty clunky to me, like a hack designed for text translation that surprised everyone by getting quite smart. The reason the human brain is so much more energy efficient is quite likely better design/algorithms. I was watching some video comparing brain and LLM function and almost tempted to try building something myself (https://youtu.be/3SUqBUGlyh8). I'm sure there are many more competent people looking at similar things.
I'm not too bullish on architectural gains. There are efficiencies to be had, but far closer to "+5% a year" than "+5000% in a single breakthrough".
You can try to build a novel AI architecture, at a small scale. Just be ready. This field will kick your teeth in. ML doesn't like grand ideas and isn't kind to high aspirations.
AI researchers came up with an architectural improvement that made a lot of previously impossible stuff barely possible. Then, industry ran with it. Scaling that particular trick to the limits by throwing as much raw compute and data at it as humanly possible.
You don't need to be an AI expert to know that there are probably more advances to be had and that funding foundational research is the way to get them.
It's not certain if something like JEPA would ever reach production grade AI models.
This has changed since the cost of equipment and infrastructure has been close to "free" for the large corporations running a cycle of funny-money investment in a bubble. This has allowed them sell access to the computing and models for just above operational energy costs (to the extent of increasing global energy prices), whilst offering free accounts to harvest data. No small competitor could possibly compete with that model.
The calculation for profitability (useful output for humans in a cost-benefit analysis) of the current setup is broken, and trades on our dreams of the future. Scaling computing forever simply does not work. It will never be profitable without further leaps forward in the technology--either more efficient models or hardware. By this time, the extent to the excessive investment in new data-centers will be clear.
The internet is a medium, an interconnection of autonomous computer networks
A "web" of hyperlinked HTML pages is one use of an internet
However this internet is more than a handful of popular websites incorporated as companies
Perhaps that's why it was called the "dot-com" bubble and not the "internet" bubble
You see that same pattern with AI now. Products are being provided for free or nearly free, and plenty is being spent on marketing.
That the key to success is developing a brand and user base faster than the next person can.
And that’s where AI makes things so hard. Creating a protective moat.
The minute Amazon stops being good about refunds or a cheaper delivery app comes out than DoorDash I’m gone. No loyalty.
And both of those are on the verge as they focus on profits over customer experience.
There is also a customer adoption curve of technology that lags far behind the technologist adoption curve. For example video on the Web failed a long time, until it didn't, when Youtube began to succeed. The problem became "boring" to technologists in some ways, but consumers gradually caught up.
AI is accelerating "let them eat cake" at rates never seen before in history, so I imagine the violence will follow soon after
> All this means two things to us: 1)The AI revolution will indeed be one of the biggest technology shifts in history. It will spark a generation of innovations that we can’t yet even imagine. 2) It’s going to take way longer to see those changes than we think it’s going to take right now.